Memory safety CVEs.
Published CVEs classified as memory safety, ranked by exploit odds.
-
CVE-2026-113752
Use after free in Windows Print Spooler Components allows an unauthorized attacker to execute code over an adjacent network.
-
CVE-2026-86389
Dell PowerProtect Data Domain, Feature Release versions 7.7.1.0 through 8.8, LTS2024 release versions 7.13.1.0 through 7.13.1.80, LTS2025 release versions 8.3.1.0 through 8.3.1.40, LTS2026 release versions 8.6.1.0 through 8.6.1.20, contain an Integer Overflow or Wraparound vulnerability. A high privileged attacker with remote access could potentially exploit this vulnerability, leading to Code execution.
-
CVE-2026-79955
An unauthenticated memory corruption vulnerability exists in a service accessed via the PAPI protocol in AOS-8 that could allow an unauthenticated remote attacker to conduct a denial of service attack. Successful exploitation could allow an attacker to disrupt the availability of the affected service.
-
CVE-2026-79893
Memory corruption vulnerabilities in AOS-10 could allow an unauthenticated remote attacker to conduct a denial of service attack. Successful exploitation could allow an attacker to disrupt the availability of the affected service.
-
CVE-2026-76430
A vulnerability in the Ultimate Packer for eXecutables (UPX) unpacker of ClamAV could allow an unauthenticated, remote attacker to cause a DoS condition on an affected device. This vulnerability is due to improper length checks, when unpacking the UPX files, which may result in an out-of-bounds read of heap memory. An attacker could exploit this vulnerability by submitting a crafted UPX for decode. A successful exploit could allow the attacker to cause the ClamAV decode process to terminate, resulting in a DoS condition on the affected software.
-
CVE-2026-123615
imgproxy 3.3.0 before 4.0.17 contains an out-of-bounds heap read and write in the ICO loader's BMP header reconstruction (vips_foreign_load_ico_header in vips/icoload.c) when an ICO entry declares a tiny data size, causing unvalidated header fields to be read and written past the allocated buffer. Unauthenticated remote attackers can supply a crafted ICO image to corrupt adjacent heap chunk metadata, triggering a glibc abort that crashes the worker.
-
CVE-2026-98968
In the Linux kernel, the following vulnerability has been resolved: amt: send the relay's General Query directly from the receive path An skb queued in a qdisc can outlive the tunnel it references through a raw pointer in skb->cb. For example, with igmp_qrv set to 1 on the relay a tunnel lives for 135s, so a netem delay of 180s on the amt device outlives it; when the tunnel expires and is freed, the subsequent dequeue triggers a use-after-free in amt_dev_xmit(). BUG: KASAN: slab-use-after-free in amt_dev_xmit+0x2763/0x2e20 Call Trace: amt_dev_xmit+0x2763/0x2e20 [drivers/net/amt.c:1262] dev_hard_start_xmit+0x22f/0x620 sch_direct_xmit+0x12e/0xac0 netem_dequeue+0x333/0xc50 net_tx_action+0x35c/0xa60 amt_send_igmp_gq() and amt_send_mld_gq() are only called from amt_request_handler(), inside the rcu_read_lock_bh() section of amt_rcv(). amt_request_handler() already has the tunnel the query is for: it found or created it inside that section. Queuing the query with dev_queue_xmit() only leads back into amt_dev_xmit(), which strips the Ethernet header and calls amt_send_membership_query() for that tunnel. Make that call directly from the two senders instead, the same way amt_send_advertisement() transmits from the receive path. The query never waits in a qdisc, the tunnel is only dereferenced inside the RCU section that found or created it, and nothing is stored in skb->cb, so no lookup or refcount is needed. Remove the query branch of amt_dev_xmit(), amt_skb_cb() and struct amt_skb_cb, which have no users left. Behaviour changes: - The relay's own General Queries no longer pass through the amt device's egress path: its qdisc, tc egress (clsact/tcx), the netfilter egress hook and packet taps. They are still visible as UDP on the underlay. - A query that is sent successfully is no longer counted as tx_dropped. The old query branch left through the unlock label, which counted every sent query as dropped. - A query that reaches amt_dev_xmit() on a relay from elsewhere, such as a userspace querier, is now dropped at the IGMP/MLD type switch. Before, it trusted whatever skb->cb held, and a NULL tunnel hit the WARN_ON(1).
-
CVE-2026-76421
A vulnerability in the LZX data algorithm for Microsoft Compiled HTML Help (CHM) file scanning in ClamAV could allow an unauthenticated, remote attacker to cause a DoS condition on an affected device. This vulnerability is due to improper boundary checks when applying the LZX data algorithm to CHM files during the decoder stage, which may result in buffer overflow. An attacker could exploit this vulnerability by sending a crafted CHM file for scanning. A successful exploit could allow the attacker to cause the ClamAV scanning process to terminate, resulting in a DoS condition on the affected software.
-
CVE-2026-125704
A use-after-free flaw was found in GStreamer core buffer value deserialization. When serialized buffer text contains an invalid character, an error path frees a mapped buffer before unmapping it. Applications that parse untrusted caps or structure data can reach a freed-buffer read and may crash. A network-facing GStreamer Data Protocol receiver can process such caps data without user interaction; further impact has not been established.
-
CVE-2026-57197
Asterisk is an open source private branch exchange and telephony toolkit. Prior to Asterisk 20.20.1, 21.12.3, 22.10.1, 23.4.1 and Certified Asterisk 20.7-cert11, 22.8-cert3, an unbounded `memcpy` of `OGG/Speex` packet data in `format_ogg_speex` can cause a heap buffer overflow and crash Asterisk when reading a file containing corrupted Speex audio. The attacker must create the file and make it available to Asterisk either through sufficiently elevated privileges or by convincing an Asterisk administrator to make it available. Asterisk must also be configured to play that file. This issue is fixed in Asterisk 20.20.1, 21.12.3, 22.10.1, 23.4.1 and Certified Asterisk 20.7-cert11, 22.8-cert3.
-
CVE-2026-119997
Heap-based buffer overflow in Microsoft Office Outlook allows an unauthorized attacker to execute code locally.
-
CVE-2026-111302
Heap-based buffer overflow in Microsoft Office Word allows an unauthorized attacker to execute code locally.
-
CVE-2026-110056
Heap-based buffer overflow in Microsoft Office Word allows an unauthorized attacker to execute code locally.
-
CVE-2026-110049
Heap-based buffer overflow in Microsoft Office Word allows an unauthorized attacker to execute code locally.
-
CVE-2026-113759
Use after free in Microsoft Office Word allows an unauthorized attacker to execute code locally.
-
CVE-2026-113901
A use-after-free issue was addressed with improved memory management. This issue is fixed in Safari 27.2, iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, visionOS 27.2. Processing maliciously crafted web content may lead to an unexpected process crash.
-
CVE-2026-112454
The issue was addressed with improved memory handling. This issue is fixed in iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, tvOS 27.2, visionOS 27.2, watchOS 27.2. Processing a maliciously crafted video file may lead to unexpected app termination or corrupt process memory.
-
CVE-2026-106808
OpenAirInterface Core Network v2.2.1 has a stack-based buffer overflow vulnerability in the OAI CN5G UDM Nudm_UEAU Generate Authentication Data endpoint. The UDM passes attacker-controlled resynchronizationInfo.rand and auts values to conv::hex_str_to_uint8(), which converts arbitrary-length hex strings into fixed-size stack buffers without validating the destination size. An unauthenticated attacker with access to the SBI interface and a valid SUPI can submit an oversized value to overwrite adjacent stack memory, corrupt the stack canary, and terminate the UDM process.
-
CVE-2026-117946
Incorrect boundary conditions in the JavaScript Engine component. This vulnerability was fixed in Firefox 159, Firefox ESR 115.44, Firefox ESR 153.6, Thunderbird 159, and Thunderbird 153.6.
-
CVE-2026-117917
Invalid pointer in the Graphics: WebRender component. This vulnerability was fixed in Firefox 159, Firefox ESR 115.44, Firefox ESR 153.6, Thunderbird 159, and Thunderbird 153.6.
-
CVE-2026-117862
Privilege escalation due to incorrect boundary conditions in the Networking component. This vulnerability was fixed in Firefox 159, Firefox ESR 115.44, Firefox ESR 153.6, Thunderbird 159, and Thunderbird 153.6.
-
CVE-2026-112413
An out-of-bounds read was addressed with improved bounds checking. This issue is fixed in iOS 26.7.2 and iPadOS 26.7.2, iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, macOS Sequoia 15.8.2, macOS Tahoe 26.7.2, visionOS 27.2. Processing a maliciously crafted file may lead to unexpected app termination.
-
CVE-2026-116568
FreeSWITCH is a Software Defined Telecom Stack enabling the digital transformation from proprietary telecom switches to a software implementation that runs on any commodity hardware. Prior to 1.11.2, rtmp_rtmp2rtpH264() trusted the length prefix of each inbound H.264 NAL unit without checking it against the bytes remaining in the RTMP message. An oversized declared unit underflowed the unsigned remaining-byte counter and advanced the parser beyond its receive buffer, where the next length-prefix read could crash the process. The flaw affects deployments that build and load mod_rtmp and accept video call sessions, commonly on TCP port 1935; placing the call requires a directory login where auth-calls is true, while no credentials are required where auth-calls is absent or false. A remote client can terminate FreeSWITCH and every active call, but the defect supplies no write primitive and memory disclosure has not been demonstrated. This issue is fixed in version 1.11.2.
-
CVE-2026-99046
In the Linux kernel, the following vulnerability has been resolved: gve: fix XSK buffer leak on error descriptor When the error bit is set in the RX completion descriptor, the buf_state and its attached buffer should be freed. In the case of AF_XDP ZC, the XSK buffer was not freed, leading to a leak. This issue was caught by LLM.
-
CVE-2026-103582
Unbounded memory allocation and missing frame size validation in Apache NMS ActiveMQ (the .NET OpenWire client, package Apache.NMS.ActiveMQ) before 2.3.0. The OpenWire unmarshalling code trusted the frame size and the length and count fields read from the network. It did not enforce a maximum frame size, and it did not check that decoded lengths fit within the frame being read. A malicious or compromised broker, or an attacker able to intercept and modify an unencrypted connection, could send a crafted OpenWire frame with an oversized or negative length. This could cause the client to attempt very large memory allocations, leading to excessive memory consumption, an OutOfMemoryException or an uncaught exception, and a denial of service of the client application. Users are recommended to upgrade to version 2.3.0, which enforces a configurable maximum frame size and validates lengths against the size of the frame being read.
-
CVE-2026-110491
The ONVIF service on the E3 can be rendered unresponsive by sending malformed or incomplete requests to certain operations. The affected service stops accepting new connections while the rest of the device continues operating normally, and will recover automatically after a period. This can be triggered repeatedly by any unauthenticated party with network access to the device.
-
CVE-2026-125712
A flaw was found in GStreamer's gst-plugins-base RTCP SDES packet parser. A remote sender able to reach an active media session can supply a malformed SDES entry whose declared length exceeds the received packet. Missing bounds validation can cause the receiver to read beyond the packet buffer, potentially crashing the application or exposing a limited amount of process memory.
-
CVE-2026-84539
An out-of-bounds write issue was addressed with improved bounds checking. This issue is fixed in macOS Golden Gate 27, macOS Tahoe 26.7.2. An app may be able to cause unexpected system termination.
-
CVE-2026-124001
NASA JPL ION-DTN 2.2.1b through 4.2.0 contains an integer overflow vulnerability in the Datagram Retransmission (DGR) convergence layer that allows attackers to corrupt heap memory by sending a maliciously crafted 64-bit svcDataLength value that wraps a 32-bit allocation size calculation. Attackers can exploit the dgr_receive API's unchecked memcpy operation using synthetic datagrams or DGR traffic over internal IPC or custom transports to trigger an out-of-bounds write, leading to denial of service or potential remote code execution.
-
CVE-2026-57188
Asterisk is an open source private branch exchange and telephony toolkit. Prior to Asterisk 20.20.1, 21.12.3, 22.10.1, 23.4.1 and Certified Asterisk 20.7-cert11, 22.8-cert3, an authenticated attacker may cause a stack buffer overflow in `res_xmpp` `XMPP namespace prefix` handling using carefully crafted packets. The `res_xmpp` module must be loaded, an XMPP connection must be configured in `xmpp.conf`, and the attacker must have an XMPP account on the same or a federated server. This issue is fixed in Asterisk 20.20.1, 21.12.3, 22.10.1, 23.4.1 and Certified Asterisk 20.7-cert11, 22.8-cert3.
-
CVE-2026-112606
A type confusion issue was addressed with improved memory handling. This issue is fixed in iOS 26.7.2 and iPadOS 26.7.2, iOS 27.2 and iPadOS 27.2, macOS Golden Gate 27.2, macOS Sequoia 15.8.2, macOS Tahoe 26.7.2, tvOS 27.2, visionOS 27.2, watchOS 27.2. An app may be able to cause unexpected system termination.
-
CVE-2026-128017
FrankenPHP before 1.13.4 contains a double free vulnerability in worker mode that frees SG(request_info).content_type_dup twice when a worker script ends its loop after a non-form POST request. Unauthenticated attackers can send repeated JSON POST requests until the worker loop ends, aborting the whole process when USE_ZEND_ALLOC=0 is set.
-
CVE-2026-119924
An out-of-bounds read in the TDS RPC procedure-name parser in the Babelfish for PostgreSQL extension versions 1.0-1.5, 2.1-2.15, 3.1-3.12, 4.0-4.8, 5.1-5.4 and in Amazon Aurora PostgreSQL with Babelfish might allow remote authenticated users to read adjacent process memory or cause a denial of service via a malformed TDS RPC request to the TDS port (default 1433). To remediate this issue, we recommend users upgrade to Babelfish for PostgreSQL extension versions 4.10, 5.6, 6.1 or later. As a workaround, restrict network access to the TDS port (default 1433). Customers using Amazon Aurora PostgreSQL with Babelfish should upgrade their cluster to Aurora PostgreSQL 13.9.16 (1.5.5), 13.23.3 (1.17.2), 14.6.16 (2.3.5), 14.17.8 (2.12.4), 14.18.6 (2.13.3), 14.19.5 (2.14.2), 14.20.5 (2.15.2), 14.22.3 (2.16.1), 14.23.1 (2.17.0), 15.10.8 (3.8.4), 15.12.8 (3.9.4), 15.13.6 (3.10.3), 15.14.5 (3.11.2), 15.15.5 (3.12.2), 15.17.3 (3.13.1), 15.18.1 (3.14.0), 16.8.8 (4.5.5), 16.9.6 (4.6.3), 16.10.5 (4.7.2), 16.11.5 (4.8.2), 16.13.3 (4.9.1), 16.14.1 (4.10.0), 17.4.8 (5.1.4), 17.5.6 (5.2.3), 17.6.5 (5.3.2), 17.7.5 (5.4.2), 17.9.3 (5.5.1), 17.10.1 (5.6.0), 18.3.4 (6.0.2), 18.4.1 (6.1.0) or later, and apply any pending maintenance to ensure they are on the latest patch.
-
CVE-2026-86574
Heap-based Buffer Overflow Vulnerability in JP1/Base. This issue affects JP1/Base: from 11-00 before 12-50-54.
-
CVE-2026-76939
Heap-based Buffer Overflow Vulnerability in JP1/Automatic Job Management System 3. This issue affects JP1/Automatic Job Management System 3 - Manager: from 11-00 before 13-50-31; JP1/Automatic Job Management System 3 - Agent: from 11-00 before 13-50-31.
-
CVE-2026-82118
Stack-based Buffer Overflow Vulnerability in HiRDB. This issue affects HiRDB Server Version 10: from 10-00 before 10-12-01; HiRDB Server Version 9: from 09-00 through 09-66-22; HiRDB Server with Additional Function Version 9: from 09-00 through 09-66/V; HiRDB Server Version 9(32): from 09-02 through 09-66-22; HiRDB Server with Additional Function Version 9(32): from 09-00 through 09-66-/V.
-
CVE-2026-76934
Stack-based Buffer Overflow Vulnerability in JP1/Automatic Job Management System 3. This issue affects JP1/Automatic Job Management System 3 - Manager: from 11-00 before 13-50-31; JP1/Automatic Job Management System 3 - Agent: from 11-00 before 13-50-31.
-
CVE-2026-99081
In the Linux kernel, the following vulnerability has been resolved: bnxt_en: fix DMA mapping length for padded small packets Stefan Fleischmann reported Intel IOMMU DMA Read faults on BCM57412 NetXtreme-E NICs when transmitting packets on VLAN/macvlan interfaces: DMAR: [DMA Read NO_PASID] Request device [18:00.0] fault addr 0xfc499000 [fault reason 0x06] PTE Read access is not set bnxt_en 0000:18:00.0 eno1np0: Abandoning msg {0xb4 0x41a} len: 0 due to firmware status: 0x2000001 ... NETDEV WATCHDOG: eno1np0 (bnxt_en): transmit queue 0 timed out The fault address (0xfc499000) is on an exact 4KB page boundary, pointing to a DMA read buffer overrun. In bnxt_start_xmit(), packets smaller than BNXT_MIN_PKT_SIZE (52 bytes), such as 42-byte untagged ARP frames, are padded: if (length < BNXT_MIN_PKT_SIZE) { pad = BNXT_MIN_PKT_SIZE - length; if (skb_pad(skb, pad)) goto tx_kick_pending; length = BNXT_MIN_PKT_SIZE; } mapping = dma_map_single(&pdev->dev, skb->data, len, DMA_TO_DEVICE); ... dma_unmap_len_set(tx_buf, len, len); However, 'len' was initialized earlier to skb_headlen(skb) (e.g. 42 bytes) and is left unadjusted after padding. Consequently, dma_map_single() and dma_unmap_len_set() map and track only 42 bytes. Later, the hardware TX buffer descriptor is programmed with the padded length: txbd->tx_bd_len_flags_type = cpu_to_le32(((len + pad) << TX_BD_LEN_SHIFT) | flags | TX_BD_FLAGS_PACKET_END); The NIC DMA engine is thus instructed to read 52 bytes from a region where only 42 bytes were DMA-mapped. If skb->data ends near the boundary of a 4KB page (within 'pad' bytes of the next page), the hardware DMA read overruns into the unmapped adjacent page, triggering an IOMMU fault. This issue was exposed after commit 447cbe95ebb9 ("vlan: fix skb_under_panic and races when toggling HW VLAN offload") because reserving extra VLAN headroom rounded LL_RESERVED_SPACE from 48 up to 64 bytes, shifting skb->data offsets and potentially causing small frames to land right against page boundaries. Fix this by using skb_put_padto(skb, BNXT_MIN_PKT_SIZE) in bnxt_start_xmit(), before skb_shinfo(skb)->nr_frags is sampled, because padding might linearize the skb. This ensures skb->len and skb_headlen(skb) reflect the padded size so that dma_map_single() maps the full buffer and the descriptor length is consistent. This also removes the temporary 'pad' variable and masking logic. The padding is done after the SW USO branch, otherwise bnxt_sw_udp_gso_xmit() would account the padding as UDP payload and send it in extra segments. Note that SW USO segments are still not padded: with IPv4, a segment carrying less than 10 bytes of UDP payload (small gso_size, or a short last segment) is sent as a frame shorter than BNXT_MIN_PKT_SIZE. This is a separate issue, left for a followup patch.
-
CVE-2026-124040
CivetWeb through 1.16, when built with USE_HTTP2 and enable_http2 on, contains a use-after-free vulnerability in handle_http2() that retains freed :method and :path pointers. Unauthenticated remote attackers can send a later HEADERS frame omitting these pseudo-headers to influence request method or path seen by access rules, or crash the server.
-
CVE-2026-125272
Description: When BIG-IP APM is licensed and provisioned, and a virtual server with the APM access profile is configured with a per-request policy containing a form-based Logon Page, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate. Impact: Traffic is disrupted while the TMM process restarts. This vulnerability allows a remote, unauthenticated attacker to cause a denial-of-service (DoS) on the BIG-IP system. There is no control plane exposure; this is a data plane issue only. Note: Software versions which have reached End of Technical Support (EoTS) are not evaluated.
-
CVE-2026-121812
Description: When BIG-IP APM is provisioned, and a virtual server is configured with an APM access profile and a per-request policy, and the tmm.access.policytrace database key is enabled, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate. The key is disabled by default. Impact: Traffic is disrupted while the TMM process restarts. This vulnerability allows a remote unauthenticated attacker to cause a denial-of-service (DoS) on the BIG-IP system . There is no control plane exposure; this is a data plane issue only. Note: Software versions which have reached End of Technical Support (EoTS) are not evaluated.
-
CVE-2026-59986
rabbitmq-c is a C-language AMQP client library for RabbitMQ. Prior to version 0.17.0, the bounds check (*offset = o + len) <= encoded.len overflows when o + len > SIZE_MAX on 32-bit systems. len is a uint32_t read from the AMQP wire as a BYTES/UTF8 field length. The overflow wraps the check to a small value that passes, producing a decoded amqp_bytes_t with a multi-GB .len pointing into a small frame buffer. This results in an out-of-bounds read of up to ~4GB of process memory on 32-bit systems (ARM, MIPS, x86 ILP32, embedded IoT AMQP clients). Enabling information disclosure or denial of service. Not exploitable on 64-bit systems where size_t is 64-bit. This issue has been patched in version 0.17.0.
-
CVE-2026-123151
A flaw was found in the GNU debugger (GDB). When GDB connects to a remote debug target, a malicious target can cause GDB to overflow a buffer on the stack while parsing the target's replies. This can crash GDB or potentially allow the target to execute arbitrary code in the GDB client process with the privileges of the user running GDB.
-
CVE-2026-105506
Buffer over-read in Windows Shell allows an authorized attacker to disclose information locally.
-
CVE-2026-114268
A flaw has been found in Tenda BE3L Pro 16.03.60.62. Affected by this vulnerability is the function formWifiBasicSet of the file /goform/WifiBasicSet. This manipulation of the argument security_5g causes buffer overflow. It is possible to initiate the attack remotely. The exploit has been published and may be used.
-
CVE-2026-126212
A heap-based buffer overflow vulnerability in Fortinet FortiAnalyzer 8.0.0 through 8.0.1, FortiAnalyzer 7.6.0 through 7.6.7, FortiAnalyzer 7.4.0 through 7.4.11, FortiAnalyzer Cloud 8.0.0 through 8.0.1, FortiAnalyzer Cloud 7.6.0 through 7.6.7, FortiAnalyzer Cloud 7.4.1 through 7.4.11, FortiManager 8.0.0 through 8.0.1, FortiManager 7.6.0 through 7.6.7, FortiManager 7.4.0 through 7.4.11, FortiManager Cloud 8.0.0 through 8.0.1, FortiManager Cloud 7.6.0 through 7.6.7, FortiManager Cloud 7.4.1 through 7.4.11 may allow an authenticated administrator to execute unauthorized code or commands or cause a denial of service via specially crafted JSON-RPC requests.
-
CVE-2026-126178
A heap-based buffer overflow vulnerability in Fortinet FortiAnalyzer 8.0.0 through 8.0.1, FortiAnalyzer 7.6.0 through 7.6.7, FortiAnalyzer 7.4.0 through 7.4.11, FortiAnalyzer 7.2.0 through 7.2.12, FortiAnalyzer Cloud 8.0.0 through 8.0.1, FortiAnalyzer Cloud 7.6.0 through 7.6.7, FortiAnalyzer Cloud 7.4.1 through 7.4.11, FortiAnalyzer Cloud 7.2.1 through 7.2.12, FortiManager 8.0.0 through 8.0.1, FortiManager 7.6.0 through 7.6.7, FortiManager 7.4.0 through 7.4.11, FortiManager 7.2.0 through 7.2.12, FortiManager Cloud 8.0.0 through 8.0.1, FortiManager Cloud 7.6.0 through 7.6.7, FortiManager Cloud 7.4.1 through 7.4.11, FortiManager Cloud 7.2.1 through 7.2.12 may allow an authenticated attacker with admin privileges to execute unauthorized code or commands via a crafted request.
-
CVE-2026-122167
Heap-based buffer overflow in Windows USB Driver allows an unauthorized attacker to elevate privileges with a physical attack.
-
CVE-2026-119830
Heap-based buffer overflow in Windows Storport Miniport Driver allows an unauthorized attacker to elevate privileges with a physical attack.
-
CVE-2026-119823
Heap-based buffer overflow in Windows Storport Miniport Driver allows an unauthorized attacker to execute code with a physical attack.
A list of CVEs whose flaw type is memory safety, sorted by how likely they are to be exploited (EPSS). Classification uses the CWE weakness id where present, and a small local model otherwise.